The present invention relates generally to the field of active RC filters. More specifically, the present invention is related to calibrating active RC filters by switching capacitors into and out of the feedback path.
Active RC filters are used in a number of analog processing applications. A typical one-pole active RC filter is illustrated in FIG. 1a. Differential active RC filter 100 comprises an operational amplifier (op-amp) 106, resistors 108a and 108b and capacitors 102a and 102b located in the feedback paths of op-amp 106. Differential active RC filter 100 also comprises resistors 104a and 104b located in the input paths of op-amp 106. To provide a differential active RC filter, a balanced amplifier is utilized. This amplifier is a double-input, double-output operational amplifier. For amplifiers of this type, the input signals Vinp and Vinn are required to be substantially equal in magnitude and opposite in sign when compared to the same reference voltage that defines the outputs of active filter 100. The reference voltage defining the outputs is the signal ground potential, which is the same, or different, as the power supply ground, depending upon the circuit configuration in which active filter 100 is utilized. The use of such amplifiers has been found to provide a number of advantages, such as a reduction of total harmonic distortion, even when large device mismatches in the input transistors or in the values of the integrating capacitors 102a and 102b exist.
For low to medium ranges of frequencies, op-amp 106 has near infinite impedance and near infinite gain. As such, op-amp 106 draws a small amount of current, if any, and nearly the entirety of the input signal for these frequencies is applied to the reactance components 102a and 102b and resistors 108a and 108b. Because of this, for active RC filters, the time constant, and hence the frequency characteristics, is almost purely a function of the reactance components and resistors of the active RC filter.
FIG. 1b illustrates another example of an active RC filter. Shown is a biquad active RC filter 120 which is a second order (two-pole) filter. Similar to one-pole active RC filter 100, the frequency characteristics of biquad filter 120 are almost completely a function of reactance components 110a, 110b, 114a and 114b and resistors 112a, 112b, 116a and 116b. 
Often, it is desirable to fabricate these active RC filters as monolithic integrated circuits. Consequently, during fabrication, the components of the filter are subject to process variations. In addition, during use, the components of the filter are subject to temperature and power supply variations. Because of these variations, the time constant, and hence the frequency characteristics, of an active RC filter can vary greatly from the nominal design value. One solution to these variations has been to utilize integrated amplifiers with external, precision components. However, this solution increases the size of the device and increases the cost of producing each device.
A number of other methods have been utilized to ameliorate the process, temperature and power supply variations which ultimately result in a variation of frequency characteristics. One such method is to place triode-mode (so as to act as variable resistors) MOS devices in series with the resistances in the active filter and control the gate of the MOS devices to vary the resistance to get the desired frequency characteristics. Another method involves replacing the resistors with an array of resistors which are switched in and out of the circuit to adjust the frequency characteristics. However, the use of the MOS devices increases non-linearities in the circuit. Resistors are more likely than other components to suffer from wide variations in value, and as such, there is difficulty in utilizing the array of resistors to properly set the frequency characteristics.
Yet another method to ameliorate the effects of variations is illustrated in FIG. 2. In this method, the capacitors are replaced with an array of capacitors. As illustrated, active filter 200 comprises an op-amp 206, resistors 204a and 204b and arrays of capacitors 202a and 202b. For each of the capacitors of capacitor arrays 202a and 202b, one terminal of the capacitor is connected to the corresponding output node of op-amp 206. The second terminal of the capacitor is connected to a switch, which, when closed, connects the second terminal of the capacitor to corresponding summing junction (op-amp input) 210a, 210b, switching the capacitor into the array. To adjust the frequency characteristics of active filter 200, each capacitor of the capacitor arrays 202a and 202b is switched in or out, with the resulting total capacitance switched into the circuit determining the frequency characteristics.
The prior art solutions using MOS devices, resistor arrays and capacitor arrays, as illustrated in FIG. 2, additionally utilize a master circuit 208 to apply the appropriate voltage to the gates of the MOS devices, or to switch feedback resistors or capacitors in the arrays in and out.
In order to determine which capacitors, in the case of the implementation of FIG. 2, are switched into and out of the circuit, master circuit 208 utilizes a stable reference clock and a replica of active RC filter 200. Master circuit 208 compares the time constant of the replicated active RC filter with the stable reference clock and, from this information, is able to determine the variation of the time constant of the replicated active filter from the nominal design time constant. Because variations in devices on a single integrated chip are minimal, the variation of the replicated active filter is essentially the same as any other active RC filters on the chip. By setting up a feedback loop based upon these comparisons, the active filters on the chip can be adjusted so as to be within tolerance of the nominal design frequency characteristics.
While these methods have made it possible to tune active RC filters, there exist drawbacks to each method. As previously described, the use of MOS devices introduces increased non-linearities into the active RC filter. Resistors are more likely to have wide process variations than other components, and therefore tuning of the active filter is more difficult with resistor arrays. However, even though capacitors may be more advantageous than the other solutions, the use of capacitor arrays has certain difficulties. Although errors due to process variations can be resolved before the filter starts normal operation by switching in and out the appropriate number of capacitors, temperature and supply voltage variations can occur at any time during the operation of the filter. Therefore, it is desirable to perform corrections during the use of the active filter. However, when the active filter is in use and a capacitor is switched into the circuit directly, the summing junction (input) of the op-amp is presented with the output voltage of the op-amp (signal voltage). This generates significant distortions in the inputs and outputs of the op-amp which have long settling times.
As a result there is still a need for a device and method to track variations in the time constant of an active RC filter without distortion due to the switching of reactance elements into and out of the feedback path of the active filter.
The present invention provides for a tunable, active RC filter and method of tuning the active RC filter which prevents distortions introduced during tuning. Generally, each tuning element in the feedback loop of the tunable active RC filter comprises a reactance element, a first switch, a second switch, and a third switch. The first switch connects a first terminal of the reactance element to a summing junction at the input of the op-amp. A second switch connects the first terminal of the reactance element to a replica of the voltage present at the summing junction (input) of the op-amp, to which the first terminal of the reactance element is connectable via the first switch. A third switch, when closed, connects the first terminal of the reactance element to the second terminal of the reactance element, which is connected to the output of the op-amp.
To tune the active filter, the reactance elements are switched into or out of the feedback path of the op-amp based upon a digital signal generated by a master circuit representing the number of elements to be switched in or out. The switching in of a single reactance element is performed using the three switches so as to substantially eliminate the distortion normally caused by switching a capacitor into the feedback path of the op-amp. Prior to being switched into the feedback path, the first and second switches are open, while the third switch is closed, shorting the capacitor so that it is discharged and does not load the op-amp output. In response to the digital signal generated by the master circuit, the third switch is opened and the second switch is closed, switching the capacitor to the replica of the voltage at the input of the op-amp. This state is maintained long enough to allow the capacitor to be charged to a voltage substantially close to the voltage which would be present across the capacitor, had it previously been in the feedback path. Once the capacitor is pre-charged, the second switch is opened and the first switch is closed, connecting the capacitor to the input of the op-amp. By pre-charging the capacitor, the input of the op-amp is not presented with the output voltage of the op-amp, rather it is presented with a voltage close to what would be across the capacitor had it been in the feedback path. This substantially eliminates the distortions caused switching the capacitor into the feedback path.